France Will Check Every User's Age to Keep Under-15s Off Social Media

The harm question and the instrument question are not the same question

On 21 July 2026 France's parliament voted to bar under-15s from social media, the first European Union state to do so, with the regulator Arcom as enforcer, new accounts blocked from 1 September 2026 and existing ones closed from January 2027.⁴ The vote settles nothing about whether social media harms teenagers, because harm was never the question a ban answers. Keep two questions apart. Does time on these platforms damage adolescent mental health? And will barring under-15s reduce that damage at a cost worth paying?

The first question is genuinely unsettled, and honesty requires saying so. The best-powered correlational study, a specification-curve analysis across three datasets covering 355,358 adolescents, found the association between digital-technology use and well-being negative but tiny, explaining at most 0.4% of the variation.⁵ A separate analysis across 72 countries found no consistent link between well-being and the global roll-out of social media, and the largest United States longitudinal brain study found no drastic changes tied to digital-technology use.³ Against that sits a real precautionary case: the US Surgeon General's advisory links frequent use to depression and anxiety and reports that adolescents on social media more than three hours a day face roughly double the risk of those symptoms.⁶ The causation debate is live, and this paper does not claim to have closed it.

A policy can be aimed at a real problem and still be the wrong tool for it. Grant the harm at its highest plausible strength. The ban still has to keep the children it names off the platforms, and it still has to do that at a price the rest of the population can afford. On both counts there is now evidence, and it runs against the law France just passed.

Australia's ban removes the accounts but not the teenagers

France is not the first to try this. Australia's under-16 ban commenced on 10 December 2025, requiring platforms from Instagram and TikTok to YouTube and Snapchat to take "reasonable steps" to keep under-16s off, under penalties reaching A$49.5 million, with no penalty for the children or parents and no legally enforceable test of whether the steps actually work.⁷ The platforms removed accounts at scale. By the eSafety Commissioner's own compliance update, about 4.7 million under-16 accounts were gone by mid-January 2026, with another 310,000 by early March. Removal did not equal compliance. Some platforms let under-16s repeatedly reattempt the same age check.⁸ Removal is not the same as cessation.

Four months in, a survey commissioned by the Molly Rose Foundation, a group that favours online restrictions and so had no reason to flatter the result, found that most of the targeted teenagers were still on the platforms. The finding is an admission against interest.

Of the 12-to-15-year-olds who once held accounts, 61% still reach at least one active account on a restricted platform, and 70% of those still using the sites call it easy to get around the rule. Only about one in twenty needed a VPN. The common method was to state a different age or borrow an older friend's login.⁹ The regulator itself says it remains unclear how many under-16s still remain on the platforms, so read these shares as the best available estimate, not a settled count.⁸ The residual share is contested. The direction is not. A ban's promise is a bright line the most at-risk children cannot cross. The children crossing it here are the ones most motivated to, and the crossing costs them nothing.

The age check falls on every adult, not just the child

To keep a 14-year-old out, a platform has to establish that every user is not a 14-year-old. A ban on under-15s is therefore a verification mandate for the whole population, and that cost is certain even where the benefit is not. Australia's government-commissioned Age Assurance Technology Trial tested more than 60 solutions from 48 providers and reached a careful verdict: age assurance is "possible, but not perfect," no single solution suits every use case, none is guaranteed effective in all deployments, and the trial warned that "unnecessary data retention may occur in apparent anticipation of future regulatory needs."¹⁰

The same trial found no substantial technological barrier to it, and this paper does not claim the checks cannot be built. The point is what they cost at population scale. Every adult who wants to post now hands an age signal, often an identity document or a biometric scan, to a platform or a vendor, and the honest primary on the technology says that data may be kept longer than it should be. France's digital minister has promised that users' personal data will be protected.⁴ Australia's trial is the record of what happens when that promise meets deployment: imperfect checks, retained data, and no guarantee. The ledger reads a large, certain privacy burden on everyone against a benefit most of the targeted teenagers already evade.

The narrower tool already exists in European law

The alternative is not hypothetical, and it is not softer regulation. The European Commission's July 2025 guidelines on protecting minors under the Digital Services Act set out a concrete list of measures: disable by default the features engineered for compulsive use, naming streaks, ephemeral content, read receipts, autoplay and push notifications, and apply age assurance that is "accurate, reliable, robust, non-intrusive and non-discriminatory."¹ Britain's Children's Code has been operating this way for years: fifteen standards requiring high-privacy defaults for minors, data minimisation, and nudge techniques and geolocation switched off by default.² These change the product a minor uses rather than the door the whole public walks through.

The strongest objection to this paper's own position is that the narrower tool leans on age assurance too, so the privacy burden is not unique to a ban. That is true and it should not be hidden. The distinction is where the check sits and what it decides. Risk-based age assurance tunes a product once a user is judged likely to be a minor. A hard age gate makes proving your age the precondition of access for the entire population. The first is targeted at the harm. The second taxes the whole public to enforce a line the harmed children step over.

France has been here before. Its 2023 "digital majority" law already required platforms to refuse under-15 registration, and on 25 October 2023 the European Commission issued an opinion against it, holding that age verification on entry conflicts with the Digital Services Act when applied to platforms established in another member state.¹¹ Most of these platforms are established in Ireland, not France, and Arcom's own age-verification enforcement has been confined to operators not established elsewhere in the Union.¹² The 2026 ban re-stages a jurisdictional fight France already lost, on the same ground.

The case for drawing a bright line

The best version of the opposing case does not rest on a study. It rests on a timeline that parents lived through. Jonathan Haidt's marshalling of it is the clearest. Across multiple indicators the youth mental-health decline begins in the early 2010s, as smartphones and social media saturate adolescence. By 2015 one in five American teenage girls used social media more than 40 hours a week. By 2023 more than 46% of teenagers reported being online almost all of the time. Haidt's claim is that this is a cause and not merely a correlate, and his remedy is a rule a family can name and hold to: no social media before 16.¹³ The Surgeon General reached the same instinct from the clinic, proposing a tobacco-style warning label and reporting that more than three hours a day roughly doubles the risk of anxiety and depression symptoms.⁶ Parents and clinicians who watch this every day do not want a design nudge buried in a settings menu. They want a line, and a design regime that still lets a twelve-year-old open the app asks a great deal of a household holding that line alone.

The goal is right, and the mechanism is where it breaks. A bright line protects only if it holds, and Australia's does not: 70% of the teenagers still on the platforms found it easy to step across.⁹ A line most of the targeted children cross is not protection. It is reassurance, and reassurance is the more dangerous outcome, because a government that has "banned" social media for children has spent its capital and can tell itself the problem is handled while the design reforms that would change the product go unbuilt. Odgers, reviewing the same evidence Haidt marshals, draws the line this paper draws: stricter content and design rules and age-aware algorithms are the reasonable tools, while age-based restrictions "are unlikely to be effective in practice — or worse, could backfire given what we know about adolescent behaviour."³ Grant the harm in full, and the age gate is still the one tool the evidence singles out as the likeliest to fail.